Hey! During the update to Multiqc 1.32, we saw in ...
# multiqc
r
Hey! During the update to Multiqc 1.32, we saw in the checksums that the fasqc heatmap is removed: https://github.com/nf-core/sarek/pull/2058/files#diff-2748fb8999bddc5dcf02aa7052c9a37d7a39afb6aa24a8cbd25290caf97a4a13 is this on purpose? I didn't see it in the release notes
m
-> #CL88J906S too
p
If someone can make me a MRE then I'll look into it
So far every time I've asked, people have just decided to downgrade MultiQC instead 😅
🤣 1
m
basically in all my snapshots I noticed that these pictures were removed when I bumped MultiQC versions:
Copy code
multiqc/multiqc_plots/{pdf,png,svg}/fastqc-{,1-,2-}status-check-heatmap.{pdf,png,svg}
but the corresponding
multiqc/multiqc_data/fastqc-status-check-heatmap.txt
is still there
r
sarek is not an MRE?!
lolsob 2
p
It just needs a couple of work directories
r
ok can run the tests locally and grab the work dirs
🙏🏻 1
fastqc and multiqc only, yes?
p
yup
r
only took a thousand years to run the test profile. I am going to blame the m chip. I was going to write up that issue, but the multiqc log looks pretty bad in general: ``````
p
oof
Copy code
plot | Unable to export plot to flat images: fastqc-status-check-heatmap
This the badger?
I guess this is a problem with plot exports with emulation:
Copy code
WARNING: The requested image's platform (linux/amd64) does not match the detected host platform (linux/arm64/v8) and no specific platform was requested
Can you try with a arm64 MultiQC image? Or on an arch64 system?
r
right, but why do we have the same then on the GHA? (Trying to a) change the container now and b) reduce the test profile to educe the noise in there)
m
I had the issue on my machine and on CI, so
linux/amd64
p
Can you share the work dir @Maxime Garcia?
r
do you need anything else from that work dir?
p
ok and the plots were definitely there in v1.31?
m
Yes
I have snapshots to prove it
m
This plot is not generated by the js version of plotly, right?
p
No, it's MultiQC run time so Python
It uses Kaleido, that spins up a headless browser to render the SVG and capture PNG / JPG versions
It's really annoying, it fails a lot shakefist
It's usually system specific though (arm, singularity etc.) so not sure exactly what's the cause here
sadpanda 2
m
Yeah, I was quite puzzled as to why just 1 file got removed
m
i think you have that file in your .nftignore https://github.com/nf-core/sarek/pull/2058/files#r2569534955
m
Yeah, so I don't snapshot the content
But I should still snapshot the file name
r
when running it manually it is also not in the results folder @mashehu (see issue above)
m
Okay, sorry just an idea I had while Scrolling
r
it would have been a great fix
p
Thanks all!
For next time - running MultiQC with
--strict
is your friend for these kinds of things
In this case, when I added that flag, it fell over immediately and gave me a big traceback that led me straight to the issue
It was not to do with kaleido, so not system specific or anything - just a bug that I introduced when I added dark mode
r
TIL, didn't know about strict 🤯
shared a GIF
👌 1
p
It used to be called
--lint
a while back, but Vlad extended it to more stuff and we renamed it.
👌 1
r
When are you planning the next release? I am just wondering if I should downgrade to 1.31 for the template merge & release in sarek or if you are planing to release soon anyways and I can wait and bump to 1.33
p
Aiming to release this week
Next day or two if all goes well
🎉 1
thankyou 1